Library-based matching recognition is the key to the application of Raman spectroscopy for material composition identification, which directly affects the accuracy of matching results. For the library matching, especially for the mixture spectrum, the single matching feature can not fully reflect the similarity between the measured sample spectrum and the spectral spectrum in the library. The spectral matching needs to comprehensively consider multiple matching features. In this paper, a new spectral integration matching method is proposed by using the logistic regression mathematical model to fuse the peak matching coefficient, the non-negative least squares matching coefficient and the cosine matching coefficient. The new method takes into account both the spectral peak information and full spectrum information of the sample. The experiment results based on the Raman spectroscopy library matching of 20 kinds of amino acid mixtures show that the spectral integration matching method has lower false positive rate.